What is Mobile SR?

MobileSR aims to bring deep-learning based single image super-resolution to mobile phones. The end-product is an Android application that can apply super-resolution to both images stored in the phone gallery, and photographs taken by the camera through the application. The application will be able to process either a single image or multiple images in the background and will be able to save or share the enhanced images, as well as show the improvement in image quality.

The most important part of the project is being able to construct an on-device program that attain reasonable performance on mobile camera images, which are already high-resolution, without encountering out-of-memory issues. Recently developed frameworks such as TensorFlow Lite and techniques such as model compression/quantization has been leveraged for this purpose.